minimax-m3.1IMAGE
Cost per task: Haiku 5.5 $0.21 vs MiniMax M3.1 Flash $0.22 vs GLM-5.3-Flash $0.28
The author argues the 'open models are 10x cheaper' claim no longer holds: cost per task is $0.21 for Haiku 5.5, $0.22 for M3.1 Flash and $0.28 for GLM-5.3-Flash, and notes M3.1 Flash has a 93% hallucination rate on AA-Omniscience.
Hendrix Liu
@HendrixLiu29208Archived 7 days agoOriginally posted 2026.10.21
Hendrix Liu @HendrixLiu29208Originally posted 2026.10.21
the "open models are 10x cheaper" narrative died this week and nobody noticed. cost per task: Haiku 5.5 $0.21, MiniMax M3.1 Flash $0.22, GLM-5.3-Flash $0.28. anthropic's small model is now the cheapest and better. And MiniMax has a 93% hallucination rate on AA-Omniscience. cheap per token ≠ cheap per answer you can trust.
x.com/hendrixliu29208/status/2112651700679184405Archived 2026.10.21 12:04